Skip to main content
MCP (Model Context Protocol) lets AI agents work with PostQueen directly: they can list your integrations, schedule posts, and generate images and videos through a standardized tool-calling interface. Connect Claude, Cursor or any MCP-compatible client to your PostQueen account and ask her to run your social media in plain language.
You do not have to run or install anything. In the protocol’s own words PostQueen is the “server” and your AI client is the “client”, but she is already running at a URL. Setup is pasting that URL into your client, and for most of them it is a single command.
Claude with the PostQueen connector: it plans a five-post launch week and schedules every post to X and LinkedIn

How it works

Point a client at one URL and your agent gets 11 tools it can call. It reads their schemas, then calls them on your behalf.

Available tools

Ten of those are the same tools the in-app assistant runs. The eleventh, ask_postqueen, exists only over MCP: your client is talking to her as an agent, so it can either drive the individual tools itself or hand the whole job to her. That is why pages about the assistant say ten and pages about MCP say eleven.
Every parameter and every field each tool sends back: Tools Reference.

Authentication

There are two ways to authenticate.

API key

Get your API key from Settings > Developers > Public API in PostQueen. Use it directly in the MCP endpoint URL or as a Bearer token.

OAuth token

Building an app for other PostQueen users? Use OAuth2 to obtain tokens. OAuth tokens start with pos_ and work like API keys when sent as a Bearer token on /mcp. The /mcp/:apiKey URL form accepts API keys only.

Connecting

The server accepts two endpoint forms. Copy-paste configuration for your client: Connect your client.
Use the /mcp endpoint with your API key or OAuth token as a Bearer token:
This method supports both API keys and OAuth tokens (prefixed with pos_).
For self-hosted instances, replace https://api.postqueen.ai with your NEXT_PUBLIC_BACKEND_URL.

Quick example

Tell your AI agent something like:
“Schedule a post to X for tomorrow at 10am: Excited to announce our new feature!”
Three tool calls go by before you look up.
1

Read your channels

The agent calls integrationList to see your connected accounts (X, LinkedIn, and so on).
2

Read the platform rules

The agent calls integrationSchema with platform: "x" to learn X’s character limits, settings, and rules.
3

Schedule the post

The agent calls integrationSchedulePostTool to schedule your post in the correct format.
That is the whole loop. All of this happens automatically: you wrote one sentence, and the post is on your calendar in the format X expects.

FAQ

No key is needed for reasoning: your MCP client (Claude, Cursor and others) brings the model. But generateImageTool and the video tools run her own generation pipeline on the server, so a self-hosted backend needs OPENAI_API_KEY (and any video provider keys) for those specific tools. The other tools work without it.
PostQueen API keys do not auto-rotate, but if you regenerate one in Settings > Developers > Public API, every MCP client using the old key stops working until you update its config. Update the URL or the Authorization header in your client config and reconnect.
Your backend already serves it. The endpoint sits at /mcp (Bearer auth) and /mcp/:apiKey (key in URL). Your reverse proxy has to forward these paths and stream the response rather than buffering it: MCP on a self-hosted install.
Not today. The current tool set is read-only on integrations and write-only on posts and media: there is no getComments or replyToComment tool. She does not read incoming comments. The closest feature is Auto Actions, which can repost your own post or add a follow-up comment on it once it crosses the like milestone you set.

Next steps

Connect your client

The URL for Claude, Cursor, Codex, ChatGPT and the rest

Tools Reference

Every parameter and every field the eleven tools send back

Examples

The flows where an agent has to discover something before it can post

Limits and troubleshooting

What the tools cannot reach, and what a bad key looks like